indicated that we could be 66-percent certain 

 that the true proportion lies between the con- 

 fidence intervals shown in table 2 . Confidence 

 intervals for streambank cover would be of 

 about the same magnitude although they 

 were not calculated. 



Some difficulty was encountered in com- 

 puting confidence intervals for continuous 

 variables. Pool data, for example, resembled 

 a Poisson distribution rather than a normal 

 distribution because a large number of the 

 transects failed to intersect any pool area. 

 In order to use computational procedures de- 

 signed for normally distributed populations, 

 four successive transects were grouped to- 

 gether and treated as a single sample. 



Although this cluster approach reduced the 

 number of samples, the estimates derived still 

 fell within acceptable limits of precision (see 



Iffonows^^""^''^^"''^ intervals were computed 



CIge= p ± S- 



where 



CIgg= The confidence interval that has 

 a 66-percent probability of con- 

 taining the true value of the 

 stream characteristic measured. 



P = 



average pool width 

 average stream width 



S-= / 1 



- 9 

 X " 



- Ibid., pp. 67-68. 



Table 3. — Precision of pool estimates 



Number of Percent of Standard Confidence 

 samples stream area error range 

 Channels (n) in pool (S=) (66% level) 

 (p) ^ 



— Percent — 

 HENRY'S FORK 



Main 



17 



19.4 



3.5 



16-23 



Trib. 



13 



33.3 



11.1 



22-44 





30 



22.4 



5.4 



17-28 





MAIN FORK 



OF BEAR 



RIVER 





Main 



1.3 



10.6 



2.6 



8-13 



Trib. 













13 



10.6 



2.6 



8-13 





HAYDEN FORK 







Main 



11 



27.2 



8.1 



19-35 



Trib. 



11 



17.3 



5.2 



12-22 





22 



24.2 



5.5 



19-30 



11 



